The Business of Grid

- WatersTechnology Staff
- 01 Jun 2007

The demand for grid is so intense that the idea of running the network as a business is alluring. Charging for CPU cycles however, creates a raft of issues—namely, what happens when the fixed-income desk can't outbid the equity side of the business? Is running IT as a business really a good idea?

At the STREET#GRID event in April, Waters invited a group of industry participants to discuss these challenges. The panel, moderated by Bob Hintze of Wachovia, consisted of Dan Schwartz, vice president of global fixed income for Citi; Thomas Chippas, director, head of autobahn equity for Deutsche Bank; Barry Childe, director of high-performance computing for Barclays Capital; Mayiz Habbal, managing director at consultancy Celent; Ram Appalaraju, vice president of marketing product technology for Azul Systems; and Tom Baehr, sales and services for DataSynapse.

Moderator Bob Hintze, Wachovia:
When it comes to running grid networks as a business, one immediately thinks about two things: first, the operational aspect of running a business and second, the funding to keep the momentum to drive that business forward. What are some of the mechanisms that you are doing either operationally or from a funding point of view?Daniel Schwartz, Citi:
We've found that there is a value proposition just from converting existing applications onto a shared resource—not even dynamically provisioning resources, but just being able to drive down the unit cost of compute.

I have not found that we have had to offer people funding in order to make the argument that they were able to recognize the value.

Moderator:
Do you have a chargeback mechanism or front-end pricing or some combination of both?Schwartz:
We decided that we would break it down and make it relatively simple, so on-boarding applications is a combination of the development team and the grid technology team. Once the applications are on board, they are charged a unit cost for a certain number of CPUs over a period of time.
Moderator:
What do you do about server farm growth or replenishment?Schwartz:
I can't say that we are necessarily in the place of replenishment yet. I guess you should be expecting a certain amount of hardware to failover over a period of time. In terms of growth, we decided to reach a first plateau and we are now at capacity on our first plateau. We're now in the process of figuring out how we're going to address our next round of inventory.
Moderator:
Where do original purchase or support decisions get made? Is that a collective community decision or is that through a focalized method?Thomas Chippas, Deutsche Bank:
Probably a smidge of one and a dash of the other, in the sense that the solution is presented, but then business users need to kick in sponsorship. It is certainly a presentation of go-your-own-way, whether it cost "A," or share cost with us and cost "B." Definitely, the idea comes from the technology side of the house. But the funding and the go-ahead certainly come from the business.
Barry Childe, Barclays Capital:
Barclays Capital's approach is certainly different. Ultimately, we take the view that the compute resources are held globally by a central group. We then have a pool of resources that effectively are applied against the business requirement. As each group comes along with their expanding requirements they basically go through a service delivery team, where they are capacity-planned and managed. The relevant resources are applied against that operating level agreement. This is quite a complex world, but between the service and the production IT teams we effectively work out how the compute resources are allocated.
Moderator:
What have you seen in terms of cost allocations or chargeback models, or taking the outputs of grid server and tying them back to the business justification or some sort of ongoing usage model?Tom Baehr, DataSynapse:
Ultimately, the goal is to get to a model where the chargeback is something that is done on a utility computing basis. When you take a look at our customers as a whole, there are many who are much further along that scale than others. The key is that when customers have taken an approach where they have adopted the grid throughout the organization, they are able to get tremendous volume to scale. They can better utilize hardware and get greater usage out of hardware, but also lower their cost and improve their overall service level agreements. Customers are demanding that from us and as they become more sophisticated and more applications are put onto the grid, it is a necessity. It is something that is being driven across our entire customer base.
Moderator:
What is Azul Systems' view on running grid as a business?Ram Appalaraju, Azul Systems:
If we look at the running IT as a business, from a systems vendor perspective, it really comes down to two factors. One is the effectiveness of the asset utilization. Are you getting the full value for the money you put in and the assets that you have, creating the service and delivering the service?

The second dimension is feeding the metrics into your reporting structure and the cost structure. You can measure the actual service use by CPU, by memory—you name it; we have a whole lot of parameters built in. We provide those tools so that in the end, there are truly shared services in alignment. Not only do you drive that utilization of shared service resources all the way to the high end, but you also have the option to tweak it.

Moderator:
What is Celent's perspective in terms of pricing and cost models and how might that be pursued by individual clients and vendors?Mayiz Habbal, Celent:
I'm sure most will agree that running IT as a business within a financial firm is a difficult task to achieve. The business people are always very difficult to meet with, and to be able to tell your organization, especially on the IT side, to make it work in that fashion, it needs a lot of due diligence. Secondly, the organizations are actually silent most of the time, which basically means that they operate separately between equities and fixed income and all the other types and effects.

At the end of the day, putting an economic model around this kind of approach really depends on the infrastructure that the organization already has, and how, ultimately, they get charged back. Is it $5 per CPU hour or is it $0.50 per CPU hour? It all depends on your organization in terms of infrastructure, and not many people actually will be able to arrive at a figure that everybody would agree to.

Moderator:
Has anyone considered outsourcing this role?Schwartz:
We've certainly considered outsourcing. All things being equal, you'd hope not to outsource if you have a large institution. You hope that your datacenter, overall unit costs of the datacenter or whatever metrics that you have in your datacenter would be sufficiently cheaper internally than externally. I think when datacenters get very large it becomes a challenge.

We've found two challenges. One is provisioning, in that the organization has trouble rapidly provisioning new hardware. The second challenge is drilling down on the cost. We have yet to outsource, so it's under consideration, but without decision. I think our preference would be not to if we could hit the constraints—redundancy and cost—that outsourcing might provide us.

Chippas:
Taking a little bit of a different view, if my sole goal were to drive down nothing but cost in my business then I could probably make a very easy argument that someone on an outsource basis could provide me more for less. But when you look at the grid, the value has yet to be realized. Right now, it is a cost saver and to some extent, an operational bottleneck remover, for lack of a better word. But, the real value of grids on the trading desk is when they make the algorithmic trading decision at the moment of trade. There is a long way to go before grids really unlock any of the value there.

My issue with outsourcing is that it takes away the incentive for the folks who support my business to become more familiar, to innovate and do the research that I need in order to be uniquely positioned in the market with something somebody else doesn't have. If all I want to do is cut my cost down, then certainly, outsourcing would be the right choice. But my desire is to innovate and have a better mousetrap. I need to keep it in house so that my people become fully versed and aware of what can be achieved.

Moderator:
If grid networks are a joint marriage between the IT side and the business side, or supply and demand, who is in the driver's seat? Is it the anxiety of the business that's driving it or the anxiety of technology to provide the supply?Chippas:
I think it depends on who has answered and who you ask. Some organizations are more technology led, while others are more business led and others are a bit more collaborative. In the equity side of the business, it's the business that has had declining margins over the last few years, increasing demands in terms of speed, latency and throughput and an ever-increasing amount of complexity. The complexity stems from, for example, Regulation NMS here in the US or the Markets in Financial Instruments Directive (MiFID) in Europe. All these things converge to create problems that can't be solved by the business or by operations and technology.

Certainly, the experience at Deutsche Bank is a bit more collaborative. I think the key is that it comes down to very basic skills. It doesn't matter if you are in IT or if you are in the business. Can you clearly define the problem and the metrics for a solution? I truly believe it is that simple. However, the "how" isn't so simple. It's far more complex.

Moderator:
Does this support the idea of starting small and making your case, and then building from there?Schwartz:
Yes, I think so. I think the examples that we've seen are starting small and growing as quickly as possible to get that cost benefit. I think what it takes is an IT organization that's very innovative, very visionary, and great at selling because they've got a concept to sell. But it also takes, from an end-user perspective, somebody that recognizes that there is additional benefit to be gained by taking full advantage of the technology. It's truly a collaborative approach.
Moderator:
From the sales perspective, for those laggards that might not yet have made the investment, will they ever see the benefits if the evidence is already in front of them and across the whole industry?Baehr:
Get on board now. Many people have made that move and what we find is that as our customers have adopted or moved to a grid strategy, those that were later in the adoption seemed to want to take bigger bites faster, and either catch up or leapfrog. It's a question of taking a step into the water now and trying it out and then growing it, or diving into the deep end of the pool. We've had success with both. I think it's cultural as well as political in the organization.
Schwartz:
I'll be a little bit of a contrarian here after having paid for technology for some time now. Often times, skipping one or two generations isn't as bad as it sounds.

We've all had to muddle through the last four or five years, in some instances with systems that have not grown by leaps and bounds in terms of their capability, due to changes in markets and budgets. So, the leaps and bounds that some people have to make by coming in late are often easier if you let them bunch up and the applications themselves are built to support it. From a business perspective, I feel there is consistency in the experience of acquiring grid services and that is a tenet of the producers of hardware, software and all the components that go with it: Get to a higher level of standardization, such that the experience is the same across platforms, across hardware and software.

I will argue that unless you have a burning need right now for what I call commodity grid services and that sort of data aggregation, overnight type calculation area, there is plenty of room for you to acquire those today. But for the more intensive, difficult things, it is a viable discussion to wait until the platforms are a bit more mature. That is an option for some people—not for all—but for some.

Childe:
It's very interesting that you hear a lot about how we are going to go and build the next space shuttle. You hear very little about how we are going to go and build the next bicycle. There is a lot of focus on a big sexy product with lots of calculation and lots of power that is going to deliver results. But more often than not, one of the key things around grid is the flexibility around the reduction in operational risk that it brings to your company. There is an awful lot of noise around the really complex things we do. In the real world, there are a lot of challenges we should be meeting on a daily basis..

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact info@waterstechnology.com or view our subscription options here: https://subscriptions.waterstechnology.com/subscribe

You are currently unable to print this content. Please contact info@waterstechnology.com to find out more.

You are currently unable to copy this content. Please contact info@waterstechnology.com to find out more.

As outlined in our terms and conditions, https://www.infopro-digital.com/terms-and-conditions/subscriptions/ (point 2.4), printing is limited to a single copy.

If you would like to purchase additional rights please email info@waterstechnology.com

You may share this content using our article tools. As outlined in our terms and conditions, https://www.infopro-digital.com/terms-and-conditions/subscriptions/ (clause 2.4), an Authorised User may only make one copy of the materials for their own personal use. You must also comply with the restrictions in clause 2.5.

If you would like to purchase additional rights please email info@waterstechnology.com